Could not connect to your MCP server

[zonzonzon]

When running the new version of n8n with the MCP server, it does not work. I have identified some causes. Is it possible to configure this in Cloudron?

https://community.n8n.io/t/error-could-not-connect-to-your-mcp-server-when-integrating-external-tool-via-sse-in-ai-agent/100957/6

[girish]

@zonzonzon I read up on this a bit. It seems we have to disable gzip compression for SSE . Can you try this change - https://git.cloudron.io/platform/box/-/commit/51d1efead6438fec879b661237a893d8e65f4a57 ?

• Edit /etc/nginx/applications/<n8n-appid>/*.conf .
• search for gzip_proxied any; and add the lines in the commit (careful , don't add the '+' in front of the lines, since it's patch) .
• systemctl reload nginx

[zonzonzon]

1. search for gzip_proxied any;

2. Add this location block right after the existing location block in the NGINX configuration file:

location /mcp {
gzip off;
proxy_buffering off;
proxy_cache off;

proxy_http_version 1.1;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $remote_addr;
proxy_set_header X-Forwarded-Host $host;

proxy_pass http://ip:5678;

}

3. After adding this code, check the NGINX configuration:

nginx -t

4. If there are no errors, reload NGINX:

systemctl reload nginx

The n8n MCP server has connected and is working properly. Thank you for your guidance. @girish

[girish]

@zonzonzon are you able to try just the below instead of adding /mcp route ? This is more generic and I hope it works across all routes.

    # compression chunking logic will interfere with SSE packet flushing
    gzip_disable "text/event-stream";

    # what responses are proxied
    gzip_proxied no-cache no-store private expired auth;

[zonzonzon]

@girish I added your code snippet above, and it doesn't work. But when I add a block like mine, it works.

location /mcp {
    gzip off;
    proxy_buffering off;
    proxy_cache off;
    
    proxy_http_version 1.1;
    proxy_set_header Host $host;
    proxy_set_header X-Real-IP $remote_addr;
    proxy_set_header X-Forwarded-For $remote_addr;
    proxy_set_header X-Forwarded-Host $host;
    
    proxy_pass http://172.18.20.102:5678;
}

[basiliolp]

I also tested both solutions, and only yours worked properly.
Thank you for the help @zonzonzon

[girish]

The correct fix needs to come from n8n - https://github.com/n8n-io/n8n/issues/14546#issuecomment-2858048370 . I guess if they don't fix it upstream , the package can bundle it's own nginx as a workaround .

[zonzonzon]

@girish It seems that every time the app restarts, the nginx configuration resets to default, right? Is there any way to keep it fixed so it doesn’t change when the app restarts? Thank you.

[girish]

@zonzonzon it will be fixed in Cloudron 9 . The intermediate release (8.3.2) didn't have this fix and thus you hit the issue .

[firmansi]

@girish Any ETA for Cloudron 9?

[edwardsalfred]

Any updates on the Cloudron 9 update?